run.sh

python run_speech_recognition_seq2seq_streaming.py \
	--model_name_or_path="emilios/whisper-medium-el" \
	--dataset_name="mozilla-foundation/common_voice_11_0" \
	--dataset_name="mozilla-foundation/common_voice_11_0,google/fleurs" \
	--dataset_config_name="el,el_gr" \
	--language="greek" \
	--train_split_name="train+validation,train+validation" \
	--eval_split_name="test,-" \
	--model_index_name="Whisper Medium El Greco" \
	--text_column_name="sentence,transcription" \
	--audio_column_name="audio,audio" \
	--resume_from_checkpoint="4000" \
	--streaming="False" \
	--max_steps="5000" \
	--max_steps="9000" \
	--output_dir="./" \
	--per_device_train_batch_size="32" \
	--per_device_eval_batch_size="16" \
	--logging_steps="25" \
	--learning_rate="1e-5" \
	--warmup_steps="500" \
	--evaluation_strategy="steps" \
	--eval_steps="1000" \
	--save_strategy="steps" \
	--save_steps="1000" \
	--generation_max_length="225" \
	--length_column_name="input_length" \
	--max_duration_in_seconds="30" \
	--freeze_feature_encoder="False" \
	--report_to="tensorboard" \
	--gradient_checkpointing \
	--fp16 \
	--overwrite_output_dir \
	--do_train="True" \
	--do_eval="True" \
	--predict_with_generate \
	--do_normalize_eval \
	--use_auth_token \
	--push_to_hub